Program Information
Cone-Beam CT Reconstruction with Gravity-Induced Motion
C Shieh1*, J Barber2 , W Counter1, J Sykes2 , P Bennett1 , S Heng3 , P White3 , S Corde3 , M Jackson3 , P Keall1 , I Feain1 , (1) University of Sydney, Camperdown, NSW, (2) Western Sydney Local Health District, Blacktown, NSW, (3) Prince of Wales Private Hospital, Sydney, NSW
Presentations
SU-F-201-1 (Sunday, July 30, 2017) 2:05 PM - 3:00 PM Room: 201
Purpose: Fixed-gantry linac designs have implications for next-generation devices such as compact linacs, MRI-Linacs, or proton systems. On fixed-gantry systems, “rotational CBCT” scans are acquired under 360° horizontal subject rotation, causing gravity-induced motion-blur. The aim is to investigate the feasibility of CBCT reconstruction with gravity-induced motion for the first time.
Methods: To mimic rotational CBCT scans, fluoroscopic scans of three rabbits under anesthesia were acquired with the kV source fixed at 0° and rabbits continuously rotating at 3°/second. Conventional CBCT scans of the rabbits at different rotation angles were also acquired for comparisons. The rotational scans were reconstructed into rotation-resolved volumes at an arbitrarily chosen rotation angle (0° in this study) using a novel data-driven method that combines digital-tomosynthesis and motion estimation. An initial reconstruction was first obtained using an iterative digital-tomosynthesis algorithm with 90° projection span. The gravity-induced motion was estimated by rigidly registering the rotational CBCT projections to the forward-projections of the initial reconstruction. The registered projections were then FDK-backprojected to yield the final reconstruction. The reconstruction results were evaluated in terms of blurriness measured by visual inspection, and spatial disparities compared to the conventional CBCT scans in the thoracic region using deformable image registration.
Results: The proposed method reduced maximum gravity-induced motion-blur from 5.2 to 0.5 mm, 4.0 to 0.3 mm, and 10.2 to 2.0 mm for the three rabbits. The mean spatial disparity between the 0° rotational CBCT and the 0° conventional CBCT scans were 2.4 mm, 2.6 mm, and 6.3 mm. Motion analysis on the conventional CBCT scans indicated that rigid motion accounts for ~80% of gravity-induced motion, justifying the use of rigid registration for motion estimation.
Conclusion: The feasibility of CBCT reconstruction with gravity-induced motion due to rotation was demonstrated for the first time. This will facilitate the development of fixed-gantry based next-generation systems.
Contact Email: